Post-layout timing-driven cell placement using an accurate net length model with movable Steiner points
Amir H. Ajami, Massoud Pedram
Abstract
Amir H. Ajami, Massoud Pedram
Abstract
This paper presents a new algorithm for timing-driven cell placement using the notion of movable Steiner points that capture the net topology. The proposed algorithm improves the timing closure at the backend of the EDA design flow. Unlike conventional flows that perform placement and routing in two separate steps and use rough estimates of the net lengths during placement, our algorithm uses accurate net lengths by considering the net topologies during the Elmore delay calculation step and dynamically updates the routing during the concurrent placement of Steiner points and cells. The simultaneous placement and routing problem is formulated as a mathematical program with a small number of variables and solved by the Han-Powell method. Experimental results demonstrate the effectiveness of the new approach compared to the conventional flows.
OpenAlex reports 21 citations for this work. Citation counts describe recorded attention and do not establish research quality.
A contribution statement is not available in the OpenAlex record.
Method details are not available in the OpenAlex metadata.
Findings are not separately available in the OpenAlex metadata.
Limitations are not available in the OpenAlex metadata.
Application details are not available in the OpenAlex metadata.
This paper presents a new algorithm for timing-driven cell placement using the notion of movable Steiner points that capture the net topology. The proposed algorithm improves the timing closure at the backend of the EDA design flow. Unlike conventional flows that perform placement and routing in two separate steps and use rough estimates of the net lengths during placement, our algorithm uses accurate net lengths by considering the net topologies during the Elmore delay calculation step and dynamically updates the routing during the concurrent placement of Steiner points and cells. The simultaneous placement and routing problem is formulated as a mathematical program with a small number of variables and solved by the Han-Powell method. Experimental results demonstrate the effectiveness of the new approach compared to the conventional flows.
Key concepts: Placement, Net (polyhedron), Routing (electronic design automation), Network routing, Steiner tree problem, Computer science, Network topology, Mathematical optimization